CN107299170B - A kind of blast-melted quality robust flexible measurement method - Google Patents
A kind of blast-melted quality robust flexible measurement method Download PDFInfo
- Publication number
- CN107299170B CN107299170B CN201710679239.XA CN201710679239A CN107299170B CN 107299170 B CN107299170 B CN 107299170B CN 201710679239 A CN201710679239 A CN 201710679239A CN 107299170 B CN107299170 B CN 107299170B
- Authority
- CN
- China
- Prior art keywords
- molten iron
- blast furnace
- weight
- blast
- iron quality
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000691 measurement method Methods 0.000 title claims abstract description 12
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims abstract description 206
- 229910052742 iron Inorganic materials 0.000 claims abstract description 103
- 238000005259 measurement Methods 0.000 claims abstract description 51
- 238000000034 method Methods 0.000 claims abstract description 51
- 229910052760 oxygen Inorganic materials 0.000 claims abstract description 36
- 239000001301 oxygen Substances 0.000 claims abstract description 36
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims abstract description 34
- 230000008569 process Effects 0.000 claims abstract description 28
- 239000007789 gas Substances 0.000 claims abstract description 18
- 238000002485 combustion reaction Methods 0.000 claims abstract description 16
- 230000035699 permeability Effects 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 37
- 239000011159 matrix material Substances 0.000 claims description 24
- 210000001015 abdomen Anatomy 0.000 claims description 7
- 238000009826 distribution Methods 0.000 claims description 6
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims description 2
- 229910000831 Steel Inorganic materials 0.000 abstract description 5
- 239000010959 steel Substances 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 13
- NINIDFKCEFEMDL-UHFFFAOYSA-N Sulfur Chemical compound [S] NINIDFKCEFEMDL-UHFFFAOYSA-N 0.000 description 9
- 229910052717 sulfur Inorganic materials 0.000 description 9
- 239000011593 sulfur Substances 0.000 description 9
- 230000002159 abnormal effect Effects 0.000 description 8
- 238000003723 Smelting Methods 0.000 description 7
- 239000003245 coal Substances 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000004519 manufacturing process Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000009529 body temperature measurement Methods 0.000 description 5
- 238000001514 detection method Methods 0.000 description 5
- 238000002347 injection Methods 0.000 description 5
- 239000007924 injection Substances 0.000 description 5
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 3
- 230000002411 adverse Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 229910052710 silicon Inorganic materials 0.000 description 3
- 239000010703 silicon Substances 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000003034 coal gas Substances 0.000 description 2
- 230000006854 communication Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- XWHPIFXRKKHEKR-UHFFFAOYSA-N iron silicon Chemical compound [Si].[Fe] XWHPIFXRKKHEKR-UHFFFAOYSA-N 0.000 description 2
- 238000010238 partial least squares regression Methods 0.000 description 2
- 229910052698 phosphorus Inorganic materials 0.000 description 2
- 239000011574 phosphorus Substances 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- UFHFLCQGNIYNRP-UHFFFAOYSA-N Hydrogen Chemical compound [H][H] UFHFLCQGNIYNRP-UHFFFAOYSA-N 0.000 description 1
- 241000124033 Salix Species 0.000 description 1
- -1 Si (silicon) content Chemical compound 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000007175 bidirectional communication Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000002817 coal dust Substances 0.000 description 1
- 239000000571 coke Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 229910052739 hydrogen Inorganic materials 0.000 description 1
- 239000001257 hydrogen Substances 0.000 description 1
- 238000012880 independent component analysis Methods 0.000 description 1
- 238000012843 least square support vector machine Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005312 nonlinear dynamic Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B5/00—Making pig-iron in the blast furnace
- C21B5/006—Automatically controlling the process
-
- C—CHEMISTRY; METALLURGY
- C21—METALLURGY OF IRON
- C21B—MANUFACTURE OF IRON OR STEEL
- C21B2300/00—Process aspects
- C21B2300/04—Modeling of the process, e.g. for control purposes; CII
Landscapes
- Engineering & Computer Science (AREA)
- Chemical & Material Sciences (AREA)
- Manufacturing & Machinery (AREA)
- Materials Engineering (AREA)
- Metallurgy (AREA)
- Organic Chemistry (AREA)
- General Factory Administration (AREA)
Abstract
The present invention proposes a kind of blast-melted quality robust flexible measurement method, comprising: acquires gas flowrate in bosh, cold flow, oxygen-enriched flow, gas permeability, the oxygen enrichment percentage, theoretical temperature combustion at current time；The data of acquisition are normalized；The blast-melted quality robust soft-sensing model constructed with polynary random weight neural network is utilized, blast-melted quality robust hard measurement is carried out, obtains Si content estimated value, P content estimated value, S content estimated value, molten iron temperature estimated value.The blast-furnace body parameter that the present invention obtains real-time measurement is as the input data of model, fully consider the sequential relationship between the hysteresis characteristic of blast furnace ironmaking process and input/output variable, construct the blast-melted quality robust soft-sensing model of nonlinear auto-companding structure, the robust hard measurement for realizing the molten steel qualities parameters such as Si content, P content, S content and molten iron temperature simultaneously avoids the hysteresis quality chemically examined offline and manual operation bring uncertain.
Description
Technical Field
The invention belongs to the technical field of blast furnace smelting automatic control, and particularly relates to a robust soft measurement method for blast furnace molten iron quality.
Background
Blast furnace iron making is a very complex nonlinear dynamic process for reducing iron from iron-containing compounds such as iron ore and smelting qualified molten iron. The quality index of molten iron is the most important production index in the blast furnace ironmaking process, and directly determines the quality of subsequent steel products and the energy consumption state in the blast furnace smelting process. In order to achieve the goals of high quality, low consumption, high yield and long service life, the blast furnace ironmaking process needs to be monitored and controlled in real time. At present, the quality of molten iron is comprehensively measured by using parameters such as Si content (chemical heat), molten iron temperature (physical heat), S content, P content and the like. However, the complex and bad characteristics of the blast furnace ironmaking process, the quality parameters of the molten iron are difficult to directly measure, and the off-line test has a long time lag. In order to achieve the final purpose by better optimizing and controlling the blast furnace ironmaking process, the internal running state of the blast furnace is comprehensively and accurately monitored in real time, and the modeling of the quality parameters of the molten iron is required. The complex dynamic characteristics of iron making and the large amount of information containing various outliers data generated at the same time bring difficulties to modeling. In order to solve the problems, a data-driven multi-element blast furnace molten iron quality parameter robust soft measurement model needs to be established by adopting the controllable parameters of the blast furnace body and the molten iron quality parameter detection analysis.
Patent publication No. CN101211383A discloses a characteristic analysis and forecast method for the silicon content of blast furnace molten iron. The method comprises the steps of taking the blast furnace process parameters of a blast furnace molten iron silicon content forecasting model as input variables, adopting an improved dynamic independent component analysis method to carry out feature extraction on sample data of the input variables, eliminating correlation among production process parameters, and establishing a dynamic recursive model for forecasting the blast furnace molten iron silicon content by using a least square support vector machine algorithm of a genetic algorithm optimization model parameter.
Patent publication No. CN103320559B provides a method for forecasting the sulfur content of molten iron in a blast furnace, which takes a short-term average value of the sulfur content, a medium-term average value of the sulfur content, a long-term average value of the sulfur content, the S content of coke entering the furnace, the S content of coal powder entering the furnace and the like as input variables for forecasting the sulfur content of the molten iron, utilizes a chemical reaction process of molten iron formed in the blast furnace, combines an RBF neural network, and forecasts the sulfur content of the next molten iron, thereby obtaining better precision for forecasting the sulfur content.
Patent publication No. CN103981317A discloses a "continuous detection method of molten iron temperature at a taphole of a blast furnace based on a temperature drop model", which finally identifies the molten iron temperature at the taphole by using temperature measurement data of a thermocouple embedded at the bottom of a molten iron runner. The method solves the problems that the blast furnace molten iron temperature detection needs manual participation, the discontinuity is discontinuous, the material consumption is high, and the temperature measurement value is unstable.
The method reported in the above patent and the similar methods related to other related documents only perform prediction and soft measurement on single molten iron quality elements (such as Si content, S content, molten iron temperature, etc.), and do not perform simultaneous multi-prediction on main parameters representing the quality of blast furnace molten iron, namely Si (silicon) content, P (phosphorus) content, S (sulfur) content and molten iron temperature, so that the molten iron quality level cannot be comprehensively reflected, and the practicability is poor. Moreover, because these methods do not consider the input/output timing and the time lag relationship in the process, the established static model does not well reflect the inherent characteristics of the blast furnace smelting process. In addition, in the actual iron-making production process, the production environment is severe, faults of devices such as detection instruments and the like and the influence of other abnormal interference are detected, and the measured data comprises outliers. The methods mainly consider soft measurement of the molten iron quality parameters under an ideal furnace condition, have poor robustness, and can not inhibit the interference of outliers and accurately predict the molten iron quality parameters when modeling data contains the outliers.
The 'multi-element molten iron quality soft measurement method based on the robust random weight neural network' applied by the patent application number '201610118914.7' can forecast multi-element molten iron quality parameters and reflect the inherent characteristics of a blast furnace ironmaking process, but only solves the robust problem that the output of modeling training data contains outliers (Y-direction abnormity), but cannot solve the robust problem that the input of the training data also contains outliers (X-direction abnormity). When the input data and the output data in the modeling training data simultaneously contain outliers (namely, the X direction and the Y direction are abnormal), the robustness of the modeling method is poor. Meanwhile, it cannot solve the problem that the multiple collinearity exists in the hidden layer output to adversely affect the modeling. In summary, at present, no method for performing multi-element robust soft measurement aiming at the quality parameters (Si content, P content, S content and molten iron temperature) of molten iron in the blast furnace smelting process exists at home and abroad.
Disclosure of Invention
In order to overcome the defects of the multi-element molten iron quality parameter forecasting method in the blast furnace smelting process, the invention provides a robust soft measurement method for the quality of the blast furnace molten iron.
A blast furnace molten iron quality robust soft measurement method comprises the following steps:
step 1, collecting the gas quantity of a furnace belly, the cold air flow, the oxygen-enriched flow, the air permeability, the oxygen-enriched rate and the theoretical combustion temperature at the current moment;
step 2, normalizing the acquired data;
and 3, performing robust soft measurement on the quality of the blast furnace molten iron by using a blast furnace molten iron quality robust soft measurement model constructed by a multivariate random weight neural network to obtain an Si content estimated value, a P content estimated value, an S content estimated value and a molten iron temperature estimated value.
The blast furnace molten iron quality robust soft measurement model in the step 3 is established by the following method:
step 3-1, determining the structure and input of a blast furnace molten iron quality robust soft measurement model: selecting a multivariate random weight neural network of a nonlinear autoregressive structure as a blast furnace molten iron quality robust soft measurement model structure, and selecting six controllable variables with the maximum correlation with blast furnace molten iron quality parameters from the controllable variables in the blast furnace ironmaking process by using a typical correlation analysis method, wherein the six controllable variables comprise: the gas flow of the furnace belly, the flow of cold air, the flow of oxygen-enriched gas, the air permeability, the oxygen-enriched rate and the theoretical combustion temperature; according to the time lag relation between the dynamic characteristic of the blast furnace ironmaking process and the molten iron quality parameters, the order of the nonlinear autoregressive structure is determined to be 1, namely the input of a blast furnace molten iron quality robust soft measurement model is as follows: the measured values of the six controllable variables at the current moment, the measured values of the six controllable variables at the last moment and the measured values of the molten iron quality parameters at the last moment are output as the estimated values of the molten iron quality parameters at the current moment;
and 3-2, training a robust soft measurement model of the quality of the blast furnace molten iron.
The step 3-2 comprises:
step 3-2-1, determining relevant parameters required by model training: activating a function type g, the number L of hidden layer nodes, the maximum iteration number F, the number B of partial least square principal elements and a convergence condition E of an output weight;
3-2-2, selecting data in a certain historical time period as a robust training data set, and performing normalization processing on all variable data in the training data set; randomly generating an input weight and a threshold between an input layer and a hidden layer; computing a hidden layer output matrix H0Initial output weight β0Initial estimated value and initial residual r of model0：
Wherein, Y0Output data in a robust training dataset;
step 3-2-3, calculating the weight of each training sample participating in modeling by a Cauchy distribution weighting function according to the distribution of the residual errors
3-2-4, calculating the weight of the training sample participating in modeling by the improved Huber weight function according to the scoring matrix of the input data in the high-dimensional space
Improved Huber weight function
Wherein,the median of the sample weight corresponding to the h-th output variable, c is a scalar, and u is a variable;
step 3-2-5, calculating comprehensive weight of training sampleObtaining output weight matrix by partial least square regressionAnd a scoring matrix ThCalculating a residual error; if the output weight meets the convergence condition or exceeds the maximum iteration times, stopping training to obtain a final blast furnace molten iron quality robust soft measurement model; otherwise, repeating the steps 3-2-3 to 3-2-4.
Advantageous effects
The method takes the blast furnace body parameters obtained by real-time measurement of conventional detection equipment on the basis of an industrial field as input data of a model, fully considers the time sequence relation between the hysteresis characteristic and the input and output variables of the blast furnace smelting process, constructs a blast furnace molten iron quality robust soft measurement model with a Nonlinear Autoregressive (NARX) structure, simultaneously realizes robust soft measurement of four molten iron quality parameters including Si content, P content, S content and molten iron temperature, comprehensively describes the blast furnace molten iron quality parameters, and avoids the hysteresis of off-line test and the uncertainty caused by manual operation. The invention not only considers the influence of abnormal interference such as faults of devices such as a detector and a transmitter, abnormal outliers in the Y direction, outliers in both the X direction and the Y direction and the like on modeling in the actual iron-making production process, greatly enhances the robustness of the model, but also solves the problem of multiple collinearity of the hidden layer space, solves the problem of modeling with more high-dimensional characteristic dimension than observation sample number by respectively calculating the score matrix and the output weight matrix of the hidden layer space and the output by applying partial least squares regression (PLS) between the hidden layer space (hidden layer output) and the modeling output, and simultaneously prevents overfitting of the model. The method provided by the invention solves the problem of bad interference of outliers in Y direction and outliers in X and Y directions which are abnormal in modeling data to the model, greatly enhances the robustness of the model, simultaneously eliminates the multiple collinearity problem of hidden layer space, enables the soft measurement to be more accurate and reliable, provides better guidance for technical operators on the blast furnace site, is more beneficial to realizing the stability and the smooth operation of the blast furnace ironmaking process, and is beneficial to keeping the blast furnace ironmaking process at the level of high quality and high yield.
Drawings
The invention is further described with reference to the following figures and detailed description. The scope of the invention is not limited to the following expressions.
FIG. 1 is a diagram of a gauge configuration for a blast furnace ironmaking process according to an embodiment of the present invention;
FIG. 2 is a block diagram of a process of a robust soft measurement method for molten iron quality of a blast furnace according to an embodiment of the present invention;
FIG. 3 is a diagram of the modeling effect of robust soft measurement of molten iron quality in a blast furnace according to an embodiment of the present invention, where (a) is a comparison curve of a predicted value and an actual value of silicon content in molten iron, (b) is a comparison curve of a predicted value and an actual value of phosphorus content, (c) is a comparison curve of a predicted value and an actual value of sulfur content, and (d) is a comparison curve of a predicted value and an actual value of molten iron temperature;
in fig. 1, the numbers represent instruments and meters required for blast furnace ironmaking, and are respectively: 1-blast furnace, 2-hot blast furnace, 3-flowmeter, 4-thermometer, 5-pressure gauge, 6-hygrometer, 7-furnace belly gas quantity measuring analyzer, 8-oxygen enrichment rate measuring analyzer, 9-air permeability measuring analyzer, 10-theoretical combustion temperature measuring analyzer, 11-data acquisition device, 12-computer system;
the reference symbols used in fig. 1 represent parameters in the blast furnace ironmaking process, which are: u. of1Gas flow of furnace bosh u2Flow rate of cold air u3Oxygen enrichment flow rate u4Air permeability, u5Oxygen enrichment rate u6The theoretical combustion temperature.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings.
As shown in fig. 2, the robust soft measurement method for the quality of the blast furnace molten iron comprises the following steps:
step 1, collecting the gas quantity u of the furnace chamber at the current moment1(m3) Flow rate u of cold air2(m3Min), oxygen-enriched flow u3(m3Min), air permeability u4(m3kPa), oxygen enrichment rate u5(vol%), theoretical combustion temperature u6(℃)；
Step 2, normalizing the acquired data;
step 3, utilizing a blast furnace molten iron quality robust soft measurement model constructed by a multivariate random weight neural network to carry out blast furnace molten iron quality robust soft measurement to obtain an estimated value of Si contentP content estimationS content estimationTemperature estimation of molten iron
The blast furnace molten iron quality robust soft measurement model in the step 3 is established by the following method:
step 3-1, determining the structure and input of a blast furnace molten iron quality robust soft measurement model: selecting a multivariate random weight neural network of a nonlinear autoregressive structure as a blast furnace molten iron quality robust soft measurement model structure, and selecting six controllable variables with the maximum correlation with blast furnace molten iron quality parameters from the controllable variables in the blast furnace ironmaking process by using a typical correlation analysis method, wherein the six controllable variables comprise: gas flow u of furnace chamber1(m3) Flow rate u of cold air2(m3Min), oxygen-enriched flow u3(m3Min), air permeability u4(m3kPa), oxygen enrichment rate u5(vol%), theoretical combustion temperature u6In degrees centigrade. According to the time lag relation between the dynamic characteristic of the blast furnace ironmaking process and the molten iron quality parameters, the order of the nonlinear autoregressive structure is determined to be 1, namely the input of a blast furnace molten iron quality robust soft measurement model is as follows: and outputting the measured values of the six controllable variables at the current moment, the measured values of the six controllable variables at the last moment and the measured values of the molten iron quality parameters at the last moment as estimated values of the molten iron quality parameters at the current moment.
The input of the robust soft measurement model of the blast furnace molten iron quality comprises the following steps:
controlled variable at present time:
gas flow u of furnace chamber1(t)(m3) Flow rate u of cold air2(t)(m3Min), oxygen-enriched flow u3(t)(m3/min)、
Air permeability u4(t)(m3kPa), oxygen enrichment rate u5(t) (vol%), theoretical combustion temperature u6(t)(℃)。
Controlled variable at last moment:
gas flow u of furnace chamber1(t-1)(m3) Flow rate u of cold air2(t-1)(m3Min), oxygen-enriched flow u3(t-1)(m3/min)、
Air permeability u4(t-1)(m3kPa), oxygen enrichment rate u5(t-1) (vol%), theoretical combustion temperature u6(t-1)(℃)。
Output variable at last time:
si content y1(t-1) (%), P content y2(t-1)(％)、
S content y3(t-1) (%), molten iron temperature y4(t-1)(℃)；
The output of the robust soft measurement model of the blast furnace molten iron quality, namely the molten iron quality parameter of the current moment needing to be estimated, comprises the following steps:
estimated value of Si content
P content estimation
S content estimation
Temperature estimation of molten iron
3-2, training a blast furnace molten iron quality robust soft measurement model;
step 3-2-1, determining relevant parameters required by model training: the method comprises the following steps of activating a function type g, the number L of hidden layer nodes, the maximum iteration number F, the number B of Partial Least Squares (PLS) principal elements and a convergence condition E of an output weight.
In the present embodiment, it is preferred that,
the activation function type g is a Sigmoid function;
the number L of the hidden layer nodes is 30;
the maximum iteration number F is 50;
partial Least Squares (PLS) principal element number B10
Convergence condition E of output weight is 10-5。
3-2-2, selecting data in a certain historical time period as a robust training data set:
Z＝{(xi，yi)|xi∈Rn，yi＝Rm，i＝1，…N}，N≥L；
(including inputting dataAnd molten iron quality parameter data Y ═ Yj(t)|j＝1，2，…，m})。
All variable data in the training data set Z are normalized; randomly generating an input weight a between an input layer and a hidden layer within a certain rangeiAnd a threshold value bi，i＝1，…，L；
Calculating a hidden layer output matrix:
the initial output weight β, the initial estimate of the model and the initial residual r are calculated as follows0：
Step 3-2-3, calculating the weight of the training sample(r represents the output direction of the training data, h represents the h-th output variable): calculating the weight of each training sample participating in modeling by a Cauchy distribution weighting function according to the distribution of the residual error;
computing normalized residual vectorsThe dimension r' is consistent with that of Y,robust scale for residual error of each iteration updateWherein mean () is a median calculation function; r isihRepresents the ith residual of the h output, rhRepresents the h residual vector;
substituting the normalized residual to a Cauchy distributed weighting functionTo obtain a weight matrix corresponding to the m-dimensional output dataFurther obtaining a weight matrix corresponding to the h output variableh is 1 … m, wherein mu is mean (r)i′)，h＝1…m，i＝1…N；
ri' denotes the i-th normalized residual,representing the mean of the normalized residuals, diag () diagonal matrix creation function.
Step 3-2-4, calculating the weight of the training sample(x denotes the input direction of the training data, h denotes the h-th output variable): calculating the weight of training samples participating in modeling by an improved Huber weight function according to a scoring matrix of input data in a high-dimensional space (hidden layer output);
the scoring matrix T is calculated by Partial Least Squares (PLS)hEach score vector tihWeight of 1 … N, 1 … mCalculated from a modified Huber weight function f (u, c), in which
The median of the sample weights corresponding to the h-th output variableh＝1…m，
Where | is | · | | is the euclidean norm, c is a scalar, u is a variable, medL1(T) is from { T1，…tnL of calculation1Mean, or other operation that calculates the spatial center of the matrix.
Step 3-2-5, calculating comprehensive weight of training sampleObtaining output weight matrix by partial least square regressionAnd a scoring matrix ThCalculating a residual error; if the output weight meets the convergence condition or exceeds the maximum iteration times, stopping training to obtain a final blast furnace molten iron quality robust soft measurement model; otherwise, repeating the steps 3-2-3 to 3-2-4.
Computing weighted hidden layer output matrixAnd weighted outputTo XhAndapplying partial least squares regression (PLS) to find a scoring matrix ThAnd output weight matrix
Order toCorrecting the scoring matrix, calculating residual errors
Iterative updatingUp to the weight matrixh＝1…Each output weight in m corresponds toj is 1, 2, …, L, h is 1, 2, …, m is less than the specified convergence condition E and the number of iterations is less than the maximum number of iterations F, the final output weight is then obtained
And (3) evaluating the final blast furnace molten iron quality robust soft measurement model by adopting two evaluation indexes of Mean Square Error (MSE) and Mean Absolute Percentage Error (MAPE):
mean square error MSE:
mean absolute percent error MAPE:
in the formula, Hiβ*For the estimate of the ith sample in the test set, YiFor the true value of the ith sample in the test set, N1Is the total number of samples in the test set.
Taking one volume of the willow steel as 2600m3The iron-making blast furnace object is taken as an example, and the online robust soft measurement method for the quality of the blast furnace molten iron is applied. The present blast furnace object is installed with the following conventional measuring system, as shown in fig. 1, including: a pressure transmitter for measuring the hot air pressure of a blast furnace hot air system, a differential pressure flowmeter for measuring the flow of cold air, a balance flowmeter for measuring the flow of rich oxygen, an air humidity sensor for measuring the humidity of blast air, an infrared thermometer for measuring the temperature of hot air, a pulverized coal flowmeter for measuring the amount of pulverized coal injection, and:
furnace bosh gas quantity measuring analyzer: cold air flow, oxygen-enriched flow and pulverized coal injection quantity obtained by measuring through a conventional instrument, and blast humidity obtained by measuring through a hygrometer are analyzed and calculated to obtain a coal gas quantity parameter of a furnace bosh; the parameters of the measuring analyzer for the gas quantity in the furnace chamber are set as follows: the amount of coal gas in the furnace chamber is 1.21 × cold blast flow/60 + (2 × oxygen-enriched flow/60) + (44.8 × blast humidity (cold blast flow/60 + (oxygen-enriched flow/60))/18000) + (22.4 × hour coal injection amount × 1000 × hydrogen content of coal dust/12000);
oxygen enrichment rate measurement analyzer: oxygen-enriched flow, blast humidity and cold air flow obtained by measuring with a conventional instrument are analyzed and calculated to obtain the oxygen-enriched rate parameter of the blast furnace; the oxygen enrichment ratio measurement analyzer parameters are set as follows: oxygen enrichment rate ═ 100 ((oxygen enrichment flow rate × (0.21+ (0.29 × blast humidity/8/100)) × (cold blast flow rate/60 + (oxygen enrichment flow rate/60)) - (0.21+ (0.29 × blast humidity/8/100)));
air permeability measurement analyzer: analyzing and calculating the air permeability parameter of the blast furnace through the cold air flow, the hot air pressure and the furnace top pressure which are obtained by measuring through a conventional instrument; the parameters are set as follows: air permeability is 100 ═ cold air flow/(hot air pressure-furnace top pressure)
Theoretical combustion temperature measurement analyzer: calculating theoretical combustion temperature parameters of the blast furnace through analysis of hot air temperature, oxygen-rich flow, cold air flow, blast air humidity and coal injection quantity per hour measured by a conventional instrument; the parameters are set as follows: theoretical combustion temperature 1559+ (0.839 hot blast temperature) + (4.972 1000 oxygen enrichment flow/cold blast flow) - (6.033 blast humidity) - (3.15 coal injection per hour 1000/cold blast flow);
the structure is shown in figure 1, 1-blast furnace, 2-hot blast furnace, 3-flowmeter, 4-thermometer, 5-pressure gauge, 6-hygrometer, 7-furnace belly gas measurement analyzer, 8-oxygen enrichment rate measurement analyzer, 9-air permeability measurement analyzer, 10-theoretical combustion temperature measurement analyzer, 11-data collector, 12-computer system;
the flowmeter 3, the thermometer 4, the pressure gauge 5, the 6-hygrometer, the furnace belly gas measurement analyzer 7, the oxygen enrichment rate measurement analyzer 8, the air permeability measurement analyzer 9, the theoretical combustion temperature measurement analyzer 10 and other conventional measuring instruments are installed at each position of the blast furnace 1, and the data acquisition 11 is connected with the conventional measuring instruments and is connected with a computer system through a communication bus.
The method can adopt C # high-level language to realize the software system. The software system can realize the functions of data display, query, soft measurement result display, query and the like, and can conveniently enable operators to obtain information such as soft measurement, historical trend, data analysis and the like required by the operators. In addition, the computer system is provided with OPC communication software for data bidirectional communication with the lower computer and the data acquisition device.
The invention utilizes the actual Liu-Steel No. 2 blast furnace and blast furnace ironmaking process data acquired by the conventional measuring equipment as data required by modeling, thereby not only solving the adverse effect of outliers with abnormal output on modeling in training data, but also solving the adverse effect of outliers with abnormal input and output in the training data, more accurately providing the estimated value of the quality parameter of the multi-element molten iron in a specified dynamic time interval, providing reference for the optimized operation and stable smooth operation of the blast furnace production process, and leading an ironmaking plant to obtain the maximum benefit.
In order to illustrate the superiority of the present invention, the effect of predicting the molten iron quality index of the molten iron quality parameter soft measurement system for a period of time is performed, as shown in fig. 3(a) to (d), wherein the data used are all data collected in the actual blast furnace ironmaking process of the liu steel No. 2. The training data of the method is 600 groups of historical samples, model evaluation is carried out through a test set, and it can be seen that curves of the forecast values of the quality indexes of the molten iron and the actual values of the quality indexes of the molten iron are basically fitted, the forecast error is minimum, and the accuracy is high. The robust soft measurement method for the quality parameter of the molten iron in the blast furnace ironmaking process has the advantages of high training speed, simple model structure, capability of solving the problem of the interference of abnormal outliers in the Y direction and the outliers in both the X direction and the Y direction on modeling, great enhancement of the robustness of the model and high prediction precision.
It should be understood that the detailed description of the present invention is only for illustrating the present invention and is not limited by the technical solutions described in the embodiments of the present invention, and those skilled in the art should understand that the present invention can be modified or substituted equally to achieve the same technical effects; as long as the use requirements are met, the method is within the protection scope of the invention.
Claims (1)
1. A blast furnace molten iron quality robust soft measurement method comprises the following steps:
step 1, collecting the gas quantity of a furnace belly, the cold air flow, the oxygen-enriched flow, the air permeability, the oxygen-enriched rate and the theoretical combustion temperature at the current moment;
step 2, normalizing the acquired data;
step 3, utilizing a blast furnace molten iron quality robust soft measurement model constructed by a multivariate random weight neural network to carry out blast furnace molten iron quality robust soft measurement to obtain an Si content estimation value, a P content estimation value, an S content estimation value and a molten iron temperature estimation value;
the blast furnace molten iron quality robust soft measurement model in the step 3 is established by the following method:
step 3-1, determining the structure and input of a blast furnace molten iron quality robust soft measurement model: selecting a multivariate random weight neural network of a nonlinear autoregressive structure as a blast furnace molten iron quality robust soft measurement model structure, and selecting six controllable variables with the maximum correlation with blast furnace molten iron quality parameters from the controllable variables in the blast furnace ironmaking process by using a typical correlation analysis method, wherein the six controllable variables comprise: the gas flow of the furnace belly, the flow of cold air, the flow of oxygen-enriched gas, the air permeability, the oxygen-enriched rate and the theoretical combustion temperature; according to the time lag relation between the dynamic characteristic of the blast furnace ironmaking process and the molten iron quality parameters, the order of the nonlinear autoregressive structure is determined to be 1, namely the input of a blast furnace molten iron quality robust soft measurement model is as follows: the measured values of the six controllable variables at the current moment, the measured values of the six controllable variables at the last moment and the measured values of the molten iron quality parameters at the last moment are output as the estimated values of the molten iron quality parameters at the current moment;
3-2, training a blast furnace molten iron quality robust soft measurement model;
characterized in that, the step 3-2 comprises:
step 3-2-1, determining relevant parameters required by model training: activating a function type g, the number L of hidden layer nodes, the maximum iteration number F, the number B of partial least square principal elements and a convergence condition E of an output weight;
3-2-2, selecting data in a certain historical time period as a robust training data set, and performing normalization processing on all variable data in the training data set; randomly generating an input weight and a threshold between an input layer and a hidden layer; computing a hidden layer output matrix H0Initial output weight β0Initial estimated value and initial residual r of model0：
Wherein, Y0Output data in a robust training dataset;
step 3-2-3, calculating the weight of each training sample participating in modeling by a Cauchy distribution weighting function according to the distribution of the residual errors
3-2-4, calculating the weight of the training sample participating in modeling by the improved Huber weight function according to the scoring matrix of the input data in the high-dimensional space
Improved Huber weight function
Wherein,the median of the sample weight corresponding to the h-th output variable, c is a scalar, and u is a variable;
step 3-2-5, calculating comprehensive weight of training sampleObtaining output weight matrix by partial least square regressionAnd a scoring matrix ThCalculating a residual error; if the output weight meets the convergence condition or exceeds the maximum iteration times, stopping training to obtain a final blast furnace molten iron quality robust soft measurement model; otherwise, repeating the steps 3-2-3 to 3-2-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710679239.XA CN107299170B (en) | 2017-08-10 | 2017-08-10 | A kind of blast-melted quality robust flexible measurement method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710679239.XA CN107299170B (en) | 2017-08-10 | 2017-08-10 | A kind of blast-melted quality robust flexible measurement method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107299170A CN107299170A (en) | 2017-10-27 |
| CN107299170B true CN107299170B (en) | 2019-02-05 |
Family
ID=60133492
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710679239.XA Active CN107299170B (en) | 2017-08-10 | 2017-08-10 | A kind of blast-melted quality robust flexible measurement method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107299170B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109117495B (en) * | 2018-06-22 | 2023-03-31 | 中南大学 | Robust data coordination method, device and storage medium in alumina production evaporation process |
| CN109359320B (en) * | 2018-08-20 | 2022-07-05 | 浙江大学 | Blast furnace index prediction method based on multiple sampling rate autoregressive distribution hysteresis model |
| CN109472070A (en) * | 2018-10-27 | 2019-03-15 | 北京化工大学 | A kind of flexible measurement method of the ESN furnace operation variable based on PLS |
| CN110320335B (en) * | 2019-07-19 | 2021-11-23 | 东北大学 | Multi-element robust soft measurement method for effluent quality index of sewage treatment |
| CN111652355B (en) * | 2020-06-02 | 2023-04-07 | 中南大学 | Method and device for predicting silicon content of blast furnace molten iron based on LSTM and DNN |
| CN111679584B (en) * | 2020-06-23 | 2022-05-03 | 武汉钢铁有限公司 | Regulating and controlling method and device for blast furnace smelting |
| CN112380779B (en) * | 2020-11-27 | 2023-11-24 | 东北大学 | Robust soft measurement method and system for sintering end point |
| CN117473770A (en) * | 2023-11-13 | 2024-01-30 | 建龙西林钢铁有限公司 | Intelligent management system for steel equipment based on digital twin information |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104651559A (en) * | 2015-01-30 | 2015-05-27 | 东北大学 | Blast furnace liquid iron quality online forecasting system and method based on multivariable online sequential extreme learning machine |
| CN105608492A (en) * | 2016-03-02 | 2016-05-25 | 东北大学 | Robust random-weight neural network-based molten-iron quality multi-dimensional soft measurement method |
| CN106909705A (en) * | 2016-12-22 | 2017-06-30 | 上海交通大学 | A kind of blast-melted quality prediction method and its system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001234213A (en) * | 2000-02-28 | 2001-08-28 | Nippon Steel Corp | Blast furnace operation method |
-
2017
- 2017-08-10 CN CN201710679239.XA patent/CN107299170B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104651559A (en) * | 2015-01-30 | 2015-05-27 | 东北大学 | Blast furnace liquid iron quality online forecasting system and method based on multivariable online sequential extreme learning machine |
| CN105608492A (en) * | 2016-03-02 | 2016-05-25 | 东北大学 | Robust random-weight neural network-based molten-iron quality multi-dimensional soft measurement method |
| CN106909705A (en) * | 2016-12-22 | 2017-06-30 | 上海交通大学 | A kind of blast-melted quality prediction method and its system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107299170A (en) | 2017-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107299170B (en) | A kind of blast-melted quality robust flexible measurement method | |
| CN107526927B (en) | Blast furnace molten iron quality online robust soft measurement method | |
| CN105608492B (en) | A kind of polynary molten steel quality flexible measurement method based on robust random weight neutral net | |
| Zhou et al. | Data-driven robust M-LS-SVR-based NARX modeling for estimation and control of molten iron quality indices in blast furnace ironmaking | |
| CN110066895B (en) | Stacking-based blast furnace molten iron quality interval prediction method | |
| CN105886680B (en) | A kind of blast furnace ironmaking process molten iron silicon content dynamic soft measuring system and method | |
| CN104651559B (en) | Blast furnace liquid iron quality online forecasting system and method based on multivariable online sequential extreme learning machine | |
| Zhou et al. | Data-driven predictive control of molten iron quality in blast furnace ironmaking using multi-output LS-SVR based inverse system identification | |
| CN109001979B (en) | Blast furnace molten iron quality self-adaptive robust prediction control method based on lazy learning | |
| Zhou et al. | Robust online sequential RVFLNs for data modeling of dynamic time-varying systems with application of an ironmaking blast furnace | |
| Zhou et al. | Multivariable dynamic modeling for molten iron quality using online sequential random vector functional-link networks with self-feedback connections | |
| Fang et al. | Soft sensors based on adaptive stacked polymorphic model for silicon content prediction in ironmaking process | |
| Yuan et al. | Double locally weighted principal component regression for soft sensor with sample selection under supervised latent structure | |
| Xie et al. | Robust stochastic configuration network multi-output modeling of molten iron quality in blast furnace ironmaking | |
| CN104750902B (en) | Molten steel quality dynamic soft-measuring method based on multi output support vector regression | |
| CN109935280A (en) | A kind of blast-melted quality prediction system and method based on integrated study | |
| CN105574297B (en) | Self adaptation blast furnace molten iron silicon content trend prediction method | |
| CN106709197A (en) | Molten iron silicon content predicting method based on slide window T-S fuzzy neural network model | |
| CN111915080B (en) | Raw fuel cost optimal proportioning method based on molten iron quality constraint | |
| CN105821170A (en) | Soft measuring system and method for quality indexes of multielement molten iron of blast furnace | |
| CN113761787A (en) | Blast furnace molten iron silicon content online prediction method and system based on deep migration network | |
| JP5583614B2 (en) | Blast furnace operation state control system, operation control device, blast furnace operation state control method and program | |
| CN111983140B (en) | Carbon monoxide measuring system and method for dry quenching production | |
| CN113177364A (en) | Soft measurement modeling method for temperature of blast furnace tuyere raceway | |
| Jiang et al. | Soft sensor modeling method based on target-guided related feature learning and its application |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |